Bonjour,
Dans le tableau suivant, je reprends ici les méthodes de Point et Sample et indique, quand j’ai pu les trouver, les équivalents Numpy.
Les méthodes de Point ont presque toutes un équivalent Numpy :
| Méthode | Description | Équivalent Numpy |
|---|---|---|
| add(*args) | Append a component (in-place). | np.append (not in-place) |
| at(*args) | Access to an element of the collection. FONCTIONNE SEULEMENT EN C++ | |
| clear() | Reset the collection to zero dimension. | np.array() (not in-place) |
| dot(rhs) | Compute the scalar product. | dot |
| find(val) | Find the index of a given value. | where |
| getClassName() | Accessor to the object’s name. | class |
| getDimension() | Accessor to the vector’s dimension. | shape |
| getId() | Accessor to the object’s id. | Pas d’équivalent |
| getName() | Accessor to the object’s name. | Pas d’équivalent |
| getShadowedId() | Accessor to the object’s shadowed id. | Pas d’équivalent |
| getSize() | Accessor to the vector’s dimension (or size). | shape |
| getVisibility() | Accessor to the object’s visibility state. | |
| hasName() | Test if the object is named. | Pas d’équivalent |
| hasVisibleName() | Test if the object has a distinguishable name. | Pas d’équivalent |
| isDecreasing() | Check if the components are in decreasing order. | np.all(np.diff(a) < 0) |
| isEmpty() | Tell if the collection is empty. | size |
| isIncreasing() | Check if the components are in increasing order. | np.all(np.diff(a) > 0) |
| isMonotonic() | Check if the components are in nonincreasing or nondecreasing order. | np.all(np.diff(a) < 0) or np.all(np.diff(a) > 0) |
| isNonDecreasing() | Check if the components are in nondecreasing order. | np.all(np.diff(a) >= 0) |
| isNonIncreasing() | Check if the components are in nonincreasing order. | np.all(np.diff(a) <= 0) |
| norm() | Compute the Euclidean () norm. | np.linalg.norm |
| norm1() | Compute the norm. | np.linalg.norm(ord=1) |
| normInf() | Compute the norm. | np.linalg.norm(ord=inf) |
| normSquare() | Compute the squared Euclidean norm. | dot(a) |
| normalize() | Compute the normalized vector with respect to its Euclidean norm. | a / np.linalg.norm(a) |
| normalizeSquare() | Compute the normalized vector with respect to its squared Euclidean norm. | a / a.dot(a) |
| resize(newSize) | Change the size of the collection. | reshape |
| setName(name) | Accessor to the object’s name. | Pas d’équivalent |
| setShadowedId(id) | Accessor to the object’s shadowed id. | Pas d’équivalent |
| setVisibility(visible) | Accessor to the object’s visibility state. | Pas d’équivalent |
Sample propose davantage de méthodes originales, mais les équivalents Numpy sont tout de même nombreux. La fonction np.std présente un piège : par défaut, elle n’utilise pas la version non biaisée, contrairement à OT.
| Méthode | Description | Équivalent Numpy |
|---|---|---|
| BuildFromPoint(point) | Static method for building a sample from a sequence of float. | reshape(-1, 1) |
| ImportFromCSVFile(*args) | Static method for building a sample from a CSV file. | genfromtxt |
| ImportFromTextFile(*args) | Static method for building a sample from a text file. | genfromtxt |
| add(*args) | Append a sample (in-place). | np.append (not in-place) |
| asPoint() | Accessor to the internal linear storage for 1D sample. | reshape(-1) |
| clear() | Erase all values. | np.array (not in-place) |
| computeCenteredMoment(k) | Estimate componentwise centered moments. | ((a – a.mean(axis=0))**k).mean() |
| computeCovariance() | Estimate the covariance matrix. | np.cov(a.T) |
| computeEmpiricalCDF(point[, tail]) | Estimate the empirical cumulative distribution function (ECDF). | Pas d’équivalent |
| computeKendallTau() | Estimate the Kendall coefficients matrix. | Pas d’équivalent |
| computeKurtosis() | Estimate the componentwise kurtosis (4th order centered normalized moment). | Pas d’équivalent |
| computeLinearCorrelation() | (ditch me?) | |
| computeMean() | Estimate the mean vector. | mean(axis=0) |
| computeMedian() | Estimate the componentwise medians (50%-quantiles). | median(axis=0) |
| computePearsonCorrelation() | Estimate the Pearson correlation matrix. | np.corrcoef(a.T) |
| computeQuantile*(args) | Estimate the quantile of the joint distribution underlying the sample. | Pas d’équivalent |
| computeQuantilePerComponent(*args) | Estimate the componentwise quantiles. | np.quantile(axis=0) |
| computeRange() | Compute the range per component. | np.ptp(axis=0) |
| computeRawMoment(k) | Compute the raw (non-centered) moment per component. | (a**k).mean() |
| computeSkewness() | Estimate the componentwise skewness (3rd order centered normalized moment). | Pas d’équivalent |
| computeSpearmanCorrelation() | Estimate the Spearman correlation matrix. | Pas d’équivalent |
| computeStandardDeviation() | Estimate the componentwise standard deviations. | std(axis=0, ddof=1) |
| computeVariance() | Estimate the componentwise variances. | var(axis=0, ddof=1) |
| erase(*args) | Erase point(s) at or between index(es) (in-place). | np.delete(a, (i), axis=0) (not in-place) |
| exportToCSVFile(*args) | Dump the sample to a CSV file. | tofile(filename, sep=‘,’) |
| find(point) | Get the position of a point in the sample. | np.where |
| getClassName() | Accessor to the object’s name. | Pas d’équivalent |
| getDescription() | Accessor to the componentwise description. | Pas d’équivalent |
| getDimension() | Accessor to the sample’s dimension. | shape[1] |
| getId() | Accessor to the object’s id. | Pas d’équivalent |
| getImplementation() | Accessor to the underlying implementation. | |
| getMarginal(*args) | Accessor to sample marginal(s) (column(s)). | a[:, index] |
| getMax() | Accessor to the componentwise maximum values. | max(axis=0) |
| getMin() | Accessor to the componentwise minimum values. | min(axis=0) |
| getName() | Accessor to the object’s name. | Pas d’équivalent |
| getSize() | Accessor to the sample size. | shape[0] |
| rank(*args) | Compute the sample (componentwise) ranks. | argsort(axis=0).argsort(axis=0) |
| select(indices) | Select points in a sample. | a[i] |
| setDescription(description) | Accessor to the componentwise description. | Pas d’équivalent |
| setName(name) | Accessor to the object’s name. | Pas d’équivalent |
| sort(*args) | Sort the sample. | sort(axis=0) |
| sortAccordingToAComponent(index) | Sort the sample according to the given component. | a[a[:, index].argsort()] |
| sortAccordingToAComponentInPlace(index) | Sort the sample in place according to the given component. | Not in-place |
| sortInPlace() | Sort the sample in place. | Not in-place |
| sortUnique() | Sort the sample and remove duplicate points. | np.unique(a).sort() |
| sortUniqueInPlace() | Sort the sample in place and remove duplicate points. | Not in-place |
| split(index) | Trunk the sample. | a[0:index] (not in-place) |
| stack(sample) | Stack (horizontally) the given sample to the current one (in-place). | np.concatenate (not in-place) |
Cordialement
Joseph